mirror of
https://github.com/runyanjake/olomana.git
synced 2025-10-04 21:27:29 -07:00
16 lines
926 B
Markdown
16 lines
926 B
Markdown
# ComfyUI
|
|
ComfyUI is the frontend package allowing you to mess around with Stable Diffusion models.
|
|
State of the art models include Flux.1 Dev, Flux.1 Schnell, and later versions of Stable Diffusion.
|
|
|
|
## Installation
|
|
```
|
|
docker compose down && docker system prune -af && docker compose up -d && docker logs -f comfyui
|
|
```
|
|
|
|
## GPU Requirements
|
|
Have a strong enough GPU to run these models. I've ran Flux.1 Dev on a 3080 TI with 12GB of VRAM.
|
|
Other than that consult charts for what you can run on your system.
|
|
At this time I'm not sure you can configure shared GPU memory on Ubuntu. On my Windows machine, I have 12GB of dedicated GPU memory, and an additional 32GB of shared GPU memory (this is normal memory dedicated as slow space the gpu can use) for a total of 44. Not sure that's possible to do via nvidia-smi on Ubuntu.
|
|
For now it may be the case that running ComfyUI on Windows is the better option for now.
|
|
|